DIY 250W/4ohm amplifier based on "blameless" topology, and measurements

[pma]

DIY 250W/4ohm power amplifier based on “blameless” topology

Hello all, the thread posted here by @sabristol
https://www.audiosciencereview.com/forum/index.php?threads/luxman-l-85v-integrated-amplifier.20657/

inspired me to build a new DIY amplifier functional sample. The circuit posted in the link above is called “Luxman L-85” but in fact the topology is rather the Douglas Self's “Blameless Amplifier” discussed in his book Audio Power Amplifier Design Handbook on also on his website
http://www.douglas-self.com/ampins/dipa/dipa.htm
http://www.douglas-self.com/ampins/dipa/dpafig33.gif

The original Luxman PB-1037 main amplifier circuit is more different with 2 differential stages instead of one, no use of EF VAS buffer etc.

@sabristol has made very good job in implementing improvements by Douglas Self (EF buffer before VAS etc.) and his circuit posted was a temptation to me to get even more power and less distortion from that circuit. The main change I made was to use 2 pairs of the output devices, to get more output current and power and less dependence on load impedance. I was thinking about my favorite and robust MJL21194/93 first but then decided to go for MJL3281/1302 pairs, which have even better linearity at high currents and are faster, though only very slightly weaker in SOA.

This is the complete schematics of the amplifier that I built

It was built into my prototype case, which determined the size of the PCB and also components placement and drilling. The case is 19” 4U, dimensions 450 x 415 x 180 mm. It has big side heatsinks and can accommodate two 300VA toroidal transformers, that are needed for the dual-mono 2x250W amplifier concept with long-term full-power capacity.

This is the amplifier PCB mounted on the heatsink

and this is the amplifier board in the prototype 19” 4U case (the bottom board. The top board is a CFA amp - it will be replaced soon)

There are two transformers, two rectifier-filter boards, two amplifier boards and two DC protection SSR boards inside the case. The metal case is grounded (connected to PE) but the signal grounds of the left and right channels are connected to the case through the Rth//C components, to prevent usual serious ground-loop hum issues. The design is dual-mono and the signal grounds of the left and right channels are not directly interconnected.

Two MJL3281/1302 output pairs make 250W/4ohm power possible with respect to SOA (Safe Operating Area) of the transistors. It is possible to use speaker/complex load that does not fall below 4 ohm in its impedance/frequency plot. The worst case simulation with the load that well reflects the woofer impedance shows that the SOA I/V trajectory of one output device is just at the edge.

This is the impedance response used in the simulation

and this is the SOA simulation for 1 power transistor, with dummy load impedance schematics

Interestingly enough the amp may drive purely resistive load of 2 ohm up to full output swing and still stay inside allowed SOA boundaries. It only tells that pure resistive loads are inadequate for both simulation and testing and do not reflect real-world speaker load.

Another interesting points of the schematics are the Q16 emitter follower (beta enhancer) that greatly reduces VAS distortion and increases open loop gain and all the current sources that improve PSR (ripple rejection).

Functional sample parameters

• input impedance ….. 70 kohm
• frequency range ….. 2Hz – 88kHz/-3dB
• full-power bandwidth ….. 20Hz – 20kHz
• output noise voltage A weighted ..... -84dBV(A)
• output power ….. 2x250W/4ohm for THD < 0.1% at 1kHz
• S/N at full power ..... 114dB(A)
• harmonic distortion ….. THD < 0.007% at 200W/4ohm/1kHz (see graphs)
• rise time of step response ….. 4us
• gain ….. 34.4dB
• dimensions ….. 450 x 415 x 180 mm
• weight ….. 30 kg approx.
• construction ….. dual-mono with 2 toroidal PSU transformers 300VA each

Measurements

Response to 10kHz square wave

Sine 20kHz at full power into 4ohm load

Distortion measurements - as mentioned in post #25, the soundcard originally used had H2 distortion similar as this amplifier so there happened a distortion cancellation in some measurements. These measurements have already been replaced by valid ones.

THD vs. output power into 4ohm load at 1kHz BW40kHz

THD vs. output power into 4ohm at 5kHz BW40kHz

THD 1kHz spectrum at 25W/4ohm/1kHz

CCIF IMD 19+20kHz at 56Vp-p/4ohm

Conclusion

The amp looks promising. Just one channel has been built at the moment. I will make a second board and then make some listening tests as well.

 

[pma]

10kHz square response measured with better oscilloscope

 

[DonH56]

Very cool! I was expecting much worse performance when I looked at the schematic until I finally noticed the AC feedback path in the upper left corner...

 

[Ryogo]

@pma I have a bit stupid question, but what is "rth//c" and how exactly it helps to prevent grounding issues?

 

[pma]

@pma I have a bit stupid question, but what is "rth//c" and how exactly it helps to prevent grounding issues?

Rth here is a varistor (my bad the label looks like it was a thermistor) and C is a capacitor. The parts are connected in parallel and connect the signal ground of PCBs (in fact at the input signal connector) with chassis and PE. The capacitor prevents the PCBs to be floating with respect to chassis but prevent from LF ground loops. The varistor prevents from creating of dangerous voltage between signal ground and chassis. This arrangement is very effective and you can use class I sound source like PC without creating LF ground loops and hum and buzz.

 

[Ryogo]

Rth here is a varistor (my bad the label looks like it was a thermistor) and C is a capacitor. The parts are connected in parallel and connect the signal ground of PCBs (in fact at the input signal connector) with chassis and PE. The capacitor prevents the PCBs to be floating with respect to chassis but prevent from LF ground loops. The varistor prevents from creating of dangerous voltage between signal ground and chassis. This arrangement is very effective and you can use class I sound source like PC without creating LF ground loops and hum and buzz.

Thank you for explanation.
I saw thread on diyaudio some time ago which discussed exactly this method but with a difference - they proposed R/C. This does helped me to get rid of hum (after picking correct resistance value), but after adding pre-amp hum returned (although, not as much as before).
So, the question is - using a varistor instead of resistor is more appropriate? And also, how one would know which nominals of capacitor and resistor/varistor should be used?

 

[pma]

So, the question is - using a varistor instead of resistor is more appropriate? And also, how one would know which nominals of capacitor and resistor/varistor should be used?

The method with a resistor, R//C, is almost equivalent. I use varistor for the reason it has very high resistance until the clamp voltage is reached. This excludes the resistive current which is frequency independent. You may choose the varistor type and clamp voltage to feel safe, like 30V. I use 180V, it is high, but I use it just for myself and the high faulty with high current capability is unlikely.

I use 47nF capacitor rated at 1000V. The capacitor impedance decreases with frequency

Xc = 1/(2*pi*f*C)

It makes 68 kohm for 50Hz and 339 ohm for 10 kHz. So for high frequencies the leakage current increases, for mains frequencies it is almost an open circuit. For high RF frequencies the chassis is effectively coupled with signal ground so it is not floating.

There are often suggestions for R//C like 10ohm//100nF, but 10ohm resistor is too small and still allows too high loop current to flow and the hum is usually not fixed. And, a resistor cannot protect against voltage difference between the signal ground and chassis. Diodes or Zener diodes are sometimes used, but I prefer the varistor as it can handle much higher current and energy.

 

[AnalogSteph]

Why have both C1 and a DC servo?

Output noise seems limited by input stage current noise and R2. That's the crux of using relatively low-beta devices like 2N5401s. Try using those for cascoding some BC560Cs (@Vce ~= 15-25 V) perhaps, or employing some KSA992s?
Gain also seems a bit higher than absolutely needed for a power amp (29.5 dB would still do fine), though I guess that was owing to the original project (the Luxman has a preamp stage of overall unity gain, pretty wacky).

I wonder why CCIF IMD would be dominant d4 and d5. Crossover, despite the faily healthy bias current? Some rail induction from layout perhaps?

My gut feeling tells me that you may be approaching the limits of an EF2... some more driver and VAS current might not go amiss.

Some 2nd-order compensation (TPC / TMC) would also be worth a try for sure.

 

[DonH56]

No idea if applicable in this case, many years since I did any sort of audio design, but I used a DC servo to provide much higher gain for lower output offset voltage as well as a narrow-band response I could tailor for its one specific purpose vs. the wideband AC feedback. I did include a large resistor in the integrator (across C19) so if the cap ever opened up the op-amp didn't rail but was never quite sure how needful that was. IIRC it was a nonpolar electrolytic back then while today it would be a ceramic or film cap.

One of the amps I built I decided to go all-out with current sources and bandgaps instead of resistors setting reference currents and such for biasing. It was rock-solid over PVT (process, voltage, temperature changes) but noisier from the added shot current. And a lot more complicated...

 

[pma]

@AnalogSteph : CCIF IMD is dominated by higher order odd harmonic only at large output swing. At some 40Vpp there is a usual decay with harmonic order. I guess 2 reasons - large signal output stage nonlinearity + dominant pole compensation decay of LG with frequency.
I added the DC servo to get <2mV output DC offset which is otherwise impossible. The C2 remained in place for the worst case reason, though the amp has speaker DC protection boards with SSR based on 90A MOSFETs - my own design.
I am sure that TPC/TMC would bring better HF linearity and probably more stability issues.
The higher gain is again to help stability and also to cover lower output sources.

 

[pma]

Capacitive load 47nF
Red is pure resistive load 4ohm, blue is 4ohm in parallel with a 47nF capacitor

THD vs. output power into resistive and resistive+capacitive load

 

[pma]

Measurements into complex dummy load (dummy load simulates real speaker load)

I have built a dummy load to simulate speaker impedance years ago. It simulates a simple 2-way box and uses highly nonlinear ferrite 18mH coil to simulate woofer impedance nonlinearity.

The load looks like this

This is the circuit schematics

and this is a measurement of the dummy load impedance

This load was now used instead of the usual and traditional purely resistive load and THD vs. frequency was measured at 9Vrms and 18Vrms. This would be 13.5W resp. 54W into 6ohm, which is an impedance minimum at some 130Hz.

Measurement in THD %

and THD in dB

One can see fast rise of distortion below 80Hz, which reflects high nonlinearity of the ferrite core 18mH coil, this is reflected in nonlinear load current and this again in voltage distortion at amp terminals due to its finite output impedance.

This complex load nonlinearity near resonance is shown in the following plot, which is THDN vs. amplitude with the dummy load at 70Hz. Please note the fast rise of ferrite coil nonlinearity effect above 10Vrms.

THDN vs. amplitude into dummy load at multiple frequencies

 

[pma]

Torture load test

Finally the test with the "torture" load

Torture load impedance and EPDR

THD vs. output voltage at 1kHz into 4ohm (red) and torture load (blue)

THD vs. power into torture load at 5kHz

THDN vs. frequency at 4Vrms

 

[pma]

Listening test

The 2nd channel was built and the amplifier has been completed. I am doing the first listening tests. The amp has great authority over the speakers and the bass control is similar and even better than with the class D amplifier as I described in
https://www.audiosciencereview.com/...testing-of-power-amplifiers.20464/post-676057
plus the top octave does not have the peaking of the class D.
https://www.audiosciencereview.com/...iyima_vs_250w4ramp_freqresp_dummy-png.119856/

The background is silent without noise and the sound is perfectly defined. Piano strikes are perfectly contoured. The sound is definitely not "sweet" and it is not a new "musical instrument" in the audio chain .

 

[sabristol]

Listening test

The 2nd channel was built and the amplifier has been completed. I am doing the first listening tests. The amp has great authority over the speakers and the bass control is similar and even better than with the class D amplifier as I described in
https://www.audiosciencereview.com/...testing-of-power-amplifiers.20464/post-676057
plus the top octave does not have the peaking of the class D.
The background is silent without noise and the sound is perfectly defined. Piano strikes are perfectly contoured. The sound is definitely not "sweet" and it is not a new "musical instrument" in the audio chain .

It's excellent for me in particular to read about this amp and how you've developed it. Full marks @pma

I do apologise for not commenting - I have been quickly reading your posts, I've just not got around to writing anything. I also need to follow up on the additional tests I said was going to do. Things are just a bit busy!

 

[pma]

It's excellent for me in particular to read about this amp and how you've developed it. Full marks @pma

I do apologise for not commenting - I have been quickly reading your posts, I've just not got around to writing anything. I also need to follow up on the additional tests I said was going to do. Things are just a bit busy!

Thank you for your kind words, @sabristol . It is me who is to say thanks to you again, because without you bringing this topology in ASR I would not think about trying it, though I of course have DS's book in my library.

I am looking forward your next measured (and listened) results. As a semi-retired consultant I have plenty of time to play with this hobby, which I would never have if I had a regular job.

 

[pma]

More distortion measurements with the dummy load. Below 100Hz we can see the rising effect of the ferrite coil nonlinearity with output voltage.

 

[pma]

Some more measurements made now with another system with wider bandwidth of 48kHz (Fs=96kHz)

1a. THD vs. output voltage and power into 4ohm at 1kHz

Green trace is the amplifier distortion and blue trace is the measurement system distortion limit in the setup used. We can see that the noise is up to about 1W defined by the measuring system noise.


1b. THD vs. output voltage into 4ohm and 6.8ohm

THD depends very little on load impedance


2. THD vs. frequency at 20V (100W) into 4ohm load and without any load

Green trace is with 4 ohm load and black trace is without load, "open" output


3. Frequency response into 4ohm and without load measured at 20V

Blue trace is without load and green trace is with 4ohm load. There is a difference of 0.18dB at 1kHz and the calculated output impedance is thus 0.084 ohm. This output impedance is mostly defined by the 2Rds resistance of output MOSFET SSR DC protection board. All the measurements taken with the output DC protection board.

 

[pma]

I think this DIY

project has been completed

 

[sabristol]

A perfect end to a perfect project

I’ve got some time off soo hopefully I’ll do something on this!